Image analysis – Image transformation or preprocessing – Changing the image coordinates
Reexamination Certificate
1999-11-09
2004-03-09
Mehta, Bhavesh M. (Department: 2625)
Image analysis
Image transformation or preprocessing
Changing the image coordinates
C382S218000, C382S278000, C345S606000, C358S525000
Reexamination Certificate
active
06704463
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention relates to an interpolation apparatus, an interpolation method and an image display apparatus and can be applied, for example, to a personal computer or an image display apparatus wherein output data having a sample number different from that of input data are produced by interpolation.
Conventionally, in video apparatus, the format of a video signal or image data is converted and processing for an image such as enlargement or reduction is executed by arithmetic processing by means of an interpolation apparatus.
In particular, a video signal of the NTSC (National Television System Committee) system includes 525 scanning lines for one frame while a video signal of the PAL (Phase Alternation by Line) system includes 625 scanning lines for one frame. Meanwhile, a video signal of the HDTV (High Definition Television) system includes 1,125 scanning lines for one frame.
Therefore, in order to process a different video signal system by different systems, interpolation operation processing is performed to convert the line number of a video signal for one frame to convert the format of the video signal in advance.
Also image data processed by a computer have different formats of various resolutions such as, for example, the VGA (Video Graphic Array) which is the pixel number of 640×480 (dots) and the SVGA (Super VGA) which is the pixel number of 800×600 (dots). Therefore, if it is tried to use, for example, a liquid crystal display panel for the SVGA to display a VGA image on the entire screen of the liquid crystal display panel, then it is necessary to convert the format of VGA image data into SVGA image data by interpolation arithmetic processing for the horizontal direction and the vertical direction.
Also in order to display an image in an enlarged or reduced scale, interpolation operation processing for the direction in which such enlargement or reduction is performed is required similarly.
Such interpolation operation processing is processing for producing image data of pixels (hereinafter referred to as conversion pixels) which form an image after conversion using image data of pixels (hereinafter referred to as original pixels) which form an original image before conversion, and produces image data of conversion pixels by convolution arithmetic operation processing by means of an interpolation filter in accordance with a positional relationship of the conversion pixels with respect to the original pixels.
In particular, taking interpolation operation processing for a horizontal direction as an example, where original pixels Ra, Rb, Rc and Rd are arranged in a juxtaposed relationship at sampling intervals S in the horizontal direction as shown in
FIG. 17
, image data at a position Q displaced by a distance x from the original pixel Rb toward the original pixel Rc side can be determined, where filter coefficients are represented by Ha, Hb, Hc and Hd, in accordance with the following expression (1):
Q=Ha×Ra+Hb×Rb+Hc×Rc+Hd×Rd
(1)
Here, theoretical ideal interpolation processing is to perform convolution operation from the infinite past to the infinite future using such a sinc function as illustrated in
FIG. 18
as an interpolation function. In
FIG. 18
, x denotes the position (distance x described hereinabove with reference to
FIG. 17
) of a conversion pixel normalized with the sampling interval S of the original pixels, and the sinc function is represented by the following expression (2):
f
(
x
)=sinc(
x
)=(sin(
x
))/
x
(2)
However, since the convolution operation from the infinite past to the infinite future is difficult, in actual interpolation operation processing, approximate operation processing is performed to produce image data of conversion pixels.
As such approximation, the most neighborhood approximation, bilinear approximation and Cubic approximation are known. Among them, the most neighborhood approximation allocates image data of the most neighboring original pixel to image date for conversion pixel and a relationship between the original pixel and the conversion pixel is represented by such an interpolation function f(x) as illustrated in FIG.
19
. In the most neighborhood approximation, the interpolation function f(x) can be represented by the following expressions (3):
f
(
x
)
=
1
(
-
0.5
<
x
≦
0.5
)
f
(
x
)
=
0
(
-
0.5
≧
x
,
x
>
0.5
)
(
3
)
Meanwhile, the bilinear approximation produces image data of a conversion pixel by weighted operation processing (weighted meaning processing) using image data of the two most neighboring pixels, and the relationship between the original pixels and the conversion pixel is represented by such an interpolation function f(x) as illustrated in FIG.
20
. In the bilinear approximation, the interpolation function f(x) can be represented by the following expressions (4):
f
(
x
)
=
1
-
&LeftBracketingBar;
x
&RightBracketingBar;
(
&LeftBracketingBar;
x
&RightBracketingBar;
≦
1
)
f
(
x
)
=
0
(
&LeftBracketingBar;
x
&RightBracketingBar;
>
1
)
(
4
)
Further, the Cubic approximation produces image data of a conversion pixel by weighted operation processing using image data of the four most neighboring pixels, and the relationship between the original pixels and the conversion pixel is represented by such an interpolation function f(x) as seen in FIG.
21
. In the Cubit approximation, the interpolation function f(x) can be represented by the following expressions (5):
f
(
x
)
=
&LeftBracketingBar;
x
&RightBracketingBar;
3
-
2
&LeftBracketingBar;
x
&RightBracketingBar;
2
+
1
(
&LeftBracketingBar;
x
&RightBracketingBar;
≦
1
)
f
(
x
)
=
-
&LeftBracketingBar;
x
&RightBracketingBar;
3
+
5
&LeftBracketingBar;
x
&RightBracketingBar;
2
-
8
&LeftBracketingBar;
x
&RightBracketingBar;
+
5
(
1
<
&LeftBracketingBar;
x
&RightBracketingBar;
≦
2
)
f
(
x
)
=
0
(
2
<
&LeftBracketingBar;
x
&RightBracketingBar;
)
(
5
)
It is to be noted that, also in
FIGS. 19
to
21
, x represents the position (distance x described hereinabove with reference to
FIG. 17
) of a conversion pixel normalized with the sampling interval S of the original pixels.
Particularly, such an interpolation filter is formed by a FIR digital filter. For example, in production of image data of a conversion pixel by the bilinear approximation, where X=0.0 (that is, an original pixel and the conversion pixel overlap each other), the image data of the conversion pixel can be determined by weighting the image data of the overlapping original pixel with the value 1.0 and weighting the image data adjacent the image data of the original pixel with the value 0.0 and then adding results of the weighting.
Further, in similar production of image data of a conversion pixel, where X=0.5 (that is, the conversion pixel is positioned at the middle point between the two adjacent original pixels), the image data of the conversion pixel can be determined by weighting the image data of the adjacent original pixels both with the value 0.5 and adding resulting values.
Furthermore, where X=0.3, the image data of the conversion pixel can be determined by weighting the image data of the most neighboring original pixel with the value 0.7 and weighting the image data of the next most neighboring pixel with the value 0.3 and then adding results of the weighting.
On the other hand, where the Cubic approximation is employed, a similar FIR digital filter is used and, where X=0.0, the image data of the four successive original pixels are weighted added with the values 0.0, 1.0, 0.0, 0.0. Consequently, the image data of that one of the four original pixels which overlaps with the conversion pixel is outputted as it is.
Where X=0.5, the image data of the four successive original pixels are weighted added with the values −0.125, 0.625, 0.625, −0.125 to produce image data of the conversion pixel
Aoyama Koji
Okada Hidehiko
Frommer William S.
Frommer & Lawrence & Haug LLP
Kassa Yosef
Mehta Bhavesh M.
Polito Bruno
LandOfFree
Interpolation/decimation apparatus, interpolation/decimation... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Interpolation/decimation apparatus, interpolation/decimation..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Interpolation/decimation apparatus, interpolation/decimation... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3191678